2. The energy productivity indicator (primary energy supply divided
by gross domestic product) in the OECD and BRIC countries fell
from 0.32 in 1971 to 0.21 in 2005 [34]. This trend can be inter-
preted as a decoupling of energy demand and economic growth.
With empirical data positively correlating the use of ICT to eco-
nomic growth [42], there is strong evidence to suggest that ICT is
an important driver for improving energy productivity. Laitner et
al. estimate that in recent years, for each kilowatt of energy used by
ICT equipment, approximately 10 kilowatts were saved economy-
wide through productivity gains and efficiency improvements [22].
So ICT is a tool for enabling energy efficiency (typically as a side
effect of process or infrastructure optimization), and has been for
many years already. However, the recent slogan “ICT for green”
also suggests that ICT is now being used more directly for energy
conservation purposes. Examples include reducing commuting by
teleworking and helping to save energy in the home. Since about
one third1
of all electricity is consumed by households, energy
conservation is an important issue.
However, while industrial processes and public infrastructures
still offer many opportunities for saving energy through conven-
tional ICT-based automation and optimization, this is more diffi-
cult in a home environment. Conventional measures for reducing
domestic energy consumption are limited – they essentially con-
sist of using more energy-efficient appliances and reducing the
energy lost by appliances on standby. Fortunately, however, ubi-
quitous computing technology (such as low-power sensors, cheap
wireless communication, embedded Web servers, etc.) is now be-
coming available, offering new opportunities for saving energy,
even without direct user involvement. Example scenarios include
automatically detecting activity in the home [44] so that the heat-
ing or air conditioning can be adjusted accordingly, and a fridge
that uses an “Internet of Things” [27] to communicate with a
smart household electricity meter in order to use any cheap excess
energy available on the power grid (for example, energy produced
by intermittent renewable energy sources) to cool itself down to
below its normal operating temperature (and thus store energy).
Although “ICT for green” alone will not save the planet, we be-
lieve that “smart” ICT can, when used consistently, reduce do-
mestic electricity consumption by at least a few percentage points.
In the rest of this paper we shall provide some arguments that
support our belief. And in any case, the “greenest” form of elec-
trical energy2
is that which doesn’t have to be produced in the first
place because it isn’t needed.
Positioning and structure of this paper. With respect to energy
conservation and ICT, we are usually concerned with the issue of
“less energy for ever more computers”. Instead, however, this pa-
per will concentrate on the dual issue of “more computers for less
energy”. The importance of the latter is motivated by a set of far-
reaching paradigm shifts in the energy sector, which we will de-
scribe in the following section. We will then discuss the potential
for ICT to favorably modify consumer demand for energy. We
outline how involving customers in ICT-driven energy conserva-
tion efforts can both foster the adoption of green products and
realize efficiency gains on top of savings from automated systems.
We then discuss how “smart” ICT can help consumers obtain im-
mediate feedback on household electricity consumption, and we
describe a prototype application (the eMeter) that aims to change
1
EU-27: 29% in 2005 [8]; U.S.: 37% in 2008 (www.eia.doe.gov/aer/).
2
With a share of 39.3% in 2007, electricity generation is the leading
source of carbon dioxide emissions in the U.S. [45].
consumer behavior by providing that feedback. We conclude the
paper with an itemization of important fields of future work.
2. PARADIGM SHIFTS IN THE ENERGY
SECTOR AND THE IMPORTANCE OF ICT
The new role of ICT as a more direct enabler of sustainable develop-
ment gives rise to a number of important challenges. These include
questions such as how technology can contribute to the optimal use
of renewable energy, how to control a changing network topology
with a huge number of energy providers, how to help establish new
energy services and solutions, and how ICT can best contribute to
smart energy marketplaces. The interest in new ICT solutions is
mainly driven by a number of (partially interwoven) paradigm
shifts within the energy sector, which we will now briefly discuss.
From “unlimited” supply to a precious resource. Building new
atomic or coal-based power plants has become unpopular in most
industrialized countries. Furthermore, the debate about the effect
of carbon dioxide on global warming and political pressure to de-
crease carbon dioxide emissions not only favors “green” energy,
but also incites us to reduce energy consumption in general. Con-
serving energy is even becoming “chic” in some circles now, and
ways of reducing energy consumption without decreasing stan-
dards of living are thus welcome.
From regulation to deregulation. Governments, particularly in
Europe, have introduced a number of measures in recent years
aimed at opening up the traditional oligopolistic and regulated
market of energy production and distribution. As new players (in-
dependent main operators, resellers, billing service providers, etc.)
enter the market, interactions across company borders are intensi-
fied. The rise of complex and timely interactions necessitates new
ICT solutions, for example to avoid costly discontinuities in
processes such as billing and to efficiently exchange the control
information necessary to operate the electrical power grid. Dere-
gulation is also leading to increased competition between players
which, together with a growing demand for green products and
services, is forcing companies to position themselves as green
players. This is even leading to the promotion of “smart energy
conservation products and services”.3
From centralized to distributed generation. Local renewable
energy generation, for example using solar panels on the roofs of
buildings, is becoming more and more important. Excess energy
that is not needed locally should be stored or fed into the grid. It is
even conceivable that in the future the batteries of parked electric
cars might act as energy buffers for transmitting power back into
the electrical grid when demand is high. Managing a bidirectional
grid and making optimal use of various small (and intermittent)
energy sources (while guaranteeing high levels of reliability) is a
non-trivial issue that requires an appropriate information and com-
munication infrastructure.
From control to cooperation. Traditionally, the electricity gener-
ated by power plants has had to match consumption at any given
point in time. In the future, power-consuming devices will increa-
singly make the most of the energy that is currently available.
More precisely, we would expect that, in a “smart grid”, energy
consuming appliances, energy generation units, power distribution
units, and various other intermediaries would negotiate and coo-
perate to optimize their situation. For this to happen, suitable ICT-
based market platforms would be required, which would then also
3
Such as the energy efficiency software company OPOWER (www.
opower.com) which helps utilities meet efficiency goals.
2
3. enable the creation of new forms of energy brokers or even virtual
power plants [38] formed by small distributed generators such as
combined heat and power (CHP) plants.
From energy consumption to smart energy usage. Many renew-
able energy sources are inferior to conventional power stations with
regard to planning and controlling energy generation. Generation
by wind turbines or photovoltaic systems in particular can lead to
major fluctuations in the energy supplied to the power grid. Such
fluctuations (and other allocation irregularities) sometimes even
lead to negative energy prices.4
This phenomenon can be miti-
gated by smart devices that consume or store energy when excess
power is available, leading to an improved balancing of supply
and demand (i.e., “demand follows generation” instead of “gener-
ation follows demand”). This not only requires an ICT infrastruc-
ture in which smart appliances can cooperate, but also measures
such as smart meters, dynamic prices, and real-time forecasting
and planning models that take account of various parameters such
as the weather, the time of day and consumption habits.
Implementing these paradigm shifts is a major undertaking that
will not only be costly and take many years, but must also be sup-
ported by the extensive use of “smart” ICT. Fortunately, recent
advances in areas such as networking, embedded systems, con-
struction automation, and ubiquitous computing can complement
conventional ICT in this respect.
Although “smart” ICT can often operate in the background and
conserve energy by optimizing and automating some processes
even in households (which are typically complex and individua-
lized environments), additional energy savings in such environ-
ments do require some consumer involvement. This represents a
challenge, since human interaction is typically regarded as inconve-
nient, and saving energy is often seen as more of a necessary con-
straint than a key objective. Nevertheless, providing consumers with
feedback on the energy consumption of their various activities and
appliances should motivate some of them to change their habits and
thus help to save energy. We will discuss this issue and appropriate
concepts and technologies in more detail in the following sections.
3. USING ICT TO INDUCE BEHAVIORAL
CHANGE
While automation and energy-optimized systems will doubtless be
essential for achieving savings, the adoption of these systems and
user behavior in general will have a major influence on the de-
mand for energy. ICT can play an important role here because it
can assist individuals in making more informed decisions and re-
ward socially desirable behavior in daily life. In fact, getting users
into the loop can not only help to guide individuals when using
energy consuming devices, but also encourage them to make fa-
vorable decisions, e.g. when purchasing electrical devices, heating
systems, and energy-efficient family cars.
There are many situations in which people – despite their general
intention to protect the environment – do not take even the sim-
plest measures to reduce their energy demand. As an example, vir-
tually all PC and imaging equipment has automated power-saving
features that switch screens or CPUs to low-power mode after a
period of user inactivity. These features, however, are all too often
not activated in either home or office environments, even though
they are pre-installed on most devices. As another example, con-
sumption in identical homes, even those designed to be low-
4
For example, on the European Energy Exchange (EEX), hour contracts
showed a price of -500.02 Euro/MWh for hour 2-3 on Oct. 4, 2009.
energy dwellings, can easily differ by a factor of two or more de-
pending on the behavior of the inhabitants [6].
The existence of many unnecessary energy sinks can be mainly at-
tributed to a lack of transparency in energy consumption [3]. One
amazing example of unanticipated growth in energy demand is
that of the market success of coffeemakers (in particular small es-
presso machines) in Swiss households and offices. For conveni-
ence, these machines often keep the water or beverage hot or even
preheat cups. In Switzerland alone, these devices consume ap-
proximately 400 GWh per year in standby mode [32]. Compared
to approximately 1000 GWh per year in total for food preparation
using kitchen hobs, ovens, microwaves, and similar cooking ap-
pliances (including coffeemakers!) in the same country [33], the
additional demand is enormous and went virtually unnoticed or at
least was not attributed to the device by most owners.
Reasons for lost saving potential originate at least in part from a
lack of knowledge on personal energy consumption, the difficul-
ties people have in investigating the efficiency of their equipment,
and a rather limited motivation to modify their personal behavior.
In order to mitigate these deficiencies, the European Union initi-
ated an Action Plan for Energy Efficiency5
in 2006 which aims at
“realizing the potential which underscores the need for a paradigm
shift to change the behavioral patterns of our societies so that we
use less energy while maintaining our quality of life” [4].
In this context, high hopes are placed on smart metering infra-
structures that provide real-time information flows and enhanced
ways of managing and controlling household energy consumption.
However, a meta study of 64 pilot projects we conducted to better
understand the efficiency gains generated by smart metering and
monthly billing showed a rather gloomy picture of the saving po-
tential achieved. After eliminating studies that had methodological
weaknesses and low explanatory power6
, the meta study showed
energy savings of just 1 to 2 percent. With direct feedback (e.g.,
using in-home displays), additional savings in the order of 1 to 2
percent have been realized (see Table 1 below for a selection of
methodologically sound pilot studies with above-average efficien-
cy gains).
The typical efficiency gains clearly lag behind common expecta-
tions. It is also worth noting that many participants in the pilot
projects were reluctant to have the metering technology installed.
A rapid decline in involvement was also common shortly after the
devices were introduced, and sustainable behavioral changes were
only achieved within a small subgroup. One could therefore con-
clude that, while many people claim that saving energy is impor-
tant, their willingness to act accordingly is rather limited [39]. But
the situation is not that hopeless. A closer look at those particular
pilot studies that used advanced motivational cues (beyond prom-
ising future cost savings) shows they typically succeeded in en-
gaging a large number of users over the duration of the campaign,
achieving significantly higher energy savings.
Based on these observations, we compiled a set of proven meas-
ures for inducing behavioral change. These measures can be cate-
gorized into two groups, one supporting rational behavior (infor-
mational support), and the other leveraging somewhat irrational
5
http://ec.europa.eu/energy/action_plan_energy_efficiency/doc/
com_2006_0545_en.pdf
6
E.g., observations that only lasted less than two months, contained no
control group, were accompanied by other efficiency measures (such as
high-profile personal energy consulting campaigns), or which included
only a priori interested participants.
3
4. motivators (intrinsic motivation and social positioning). Both cat-
egories are outlined below.
Informational support. It is widely accepted that communicating
consumption data in the form of a mere value and physical unit is
not adequate for most people [26]. For a more thorough interpre-
tation, analogies are regarded as helpful, and can also increase the
timescales over which users reflect and process the information.
The type of analogy must be chosen carefully however, to guide
the user in the desired direction, e.g. specifying the size of solar
panel that is required to produce an energy equivalent, for exam-
ple, manifests the feeling that the amount of energy is large; men-
tioning the number of cups of tea that could be prepared using this
energy has the opposite effect.
Table 1. Efficiency gains reported in feedback studies
Project lead
& country
House-
holds
Energy savings Source
Mountain,
Canada
505
6.5% against baseline over 2.5
years. Adjusted for weather &
demographics.
[30]
SydEnergy,
SEAS-NVE,
Denmark
677
2-3% electricity savings. Sig-
nificant at 5% and 10% level.
ESMA,
Togeby
7
Arvola,
Finland
525
3% against controls in 2 year
study for feedback; 5%
for feedback & advices.
[1]
Henryson,
Scandinavia
600-
1500
Between 0% and 12%. [15]
Government,
Sweden
6 mil-
lions
Approximately 3%.
ESMA,
WP2D68
Nielsen,
Denmark
~1500 1% in flats, 10% in houses. [11], [31]
Hydro One,
Canada
500 Between 7% and 10%.
ESMA,
WP2D78
Wilhite/Ling,
Norway
~1000
10% against controls over
3 years.
[6], [47]
Another important way to help place personal consumption in a
wider context is by making comparisons with other entities (fami-
lies, homes, etc.). Care must be taken when choosing average val-
ues – showing individuals that they perform better than average
regularly leads to a reduction in effort and ultimately to higher
energy demand. The same effect occurs when recipients are con-
fronted with average values that are much better than their own
performance, as they often perceive the goal as being too difficult
to achieve and therefore not worth pursuing [41].
When provided appropriately, informational support increases the
willingness to act. In order to transform the momentum into
change, consumption data should be accompanied by concrete and
context-specific advice, an offer of further assistance, or at least
some request for self-commitment.
Intrinsic motivation and social positioning. While many people
agree on the importance of their personal engagement, they often
lack motivation to ultimately take action. Established consumer
research and marketing concepts appear to be promising ways of
increasing intrinsic user motivation to achieve goals such as con-
serving energy. These concepts include goal setting, the use of
virtual budgets, and social comparisons.
Goal setting theory, in brief, asserts that goals lead to more effort
and increased persistence. Important influencing variables include
attainability and self-efficacy, and also the source that defined the
goal. Goals that one sets oneself, for example, are more likely to
7
http://ea-energianalyse.dk/publications_uk.html
be achieved than those set by external sources [25]. The degree of
ambition can be positively influenced by providing appropriate de-
faults or by stating that attachment figures or authorities have made
a specific selection. Energy monitors, for example, can combine de-
faults, goal setting, and feedback on the state of current perfor-
mance, while providing advice on ways of achieving objectives.
Energy budgets appear to be a good method of increasing intrinsic
motivation. In a British pilot project, pre-paid electricity tariffs
with simple interfaces to keep track of the current balance posi-
tively influenced saving efforts [7].
Comparisons with other entities have already been outlined as in-
formational cues. They are especially effective when the selected
peer is similar to the recipient of the information, lives in close
proximity (e.g., the same village), has the same profession, or is a
member of a familiar or admired group [29]. Moreover, people
tend to act in a socially preferable way when their behavior be-
comes visible to others. Initial projects are using social networks
such as twitter and facebook as a platform for energy efficiency
activities, but have not yet gained much attention.8
“Smart” ICT makes it possible to combine informational support
and ways of fostering intrinsic motivation. In an ideal scenario,
the deployment of energy measurement devices and energy con-
servation services would be embedded within a wider campaign,
game, or competition to get users involved. Lotteries have proved
to be efficient as an initial motivator9
, but other incentives that can
be easily facilitated using ICT have not yet been tested on a larger
scale.
We will describe a prototype system and demonstrator (the
eMeter) for testing and evaluating some of the abovementioned
concepts below, after surveying and briefly discussing in the next
section the most important energy feedback systems that have
been developed in recent years.
4. FEEDBACK ON ELECTRICITY USAGE
Several energy monitoring solutions already exist that can provide
feedback on electricity consumption. They aim to help users un-
derstand where energy wastage occurs and thus try to establish a
basis for conscious energy usage. These electricity feedback solu-
tions can broadly be classified into two categories according to the
number (and type) of sensors used to acquire the electricity con-
sumption information.
4.1 Single Sensor Approach
The first category consists of single sensor solutions, which are
primarily limited to displaying the aggregated consumption of a
circuit or even the entire power demand of a household. There are
several products available, such as Wattson10
, Onzo11
, Current
Cost12
, Power Cost Monitor13
, and TED-100014
. Once installed,
they visualize the overall electricity consumption on a display
unit. However, installation at a circuit or household level is com-
plex and users are therefore often discouraged from using such
products. Furthermore, these solutions suffer from the fact that,
8
E.g., Energy Monsters, www.facebook.com/home.php?#/apps/
application.php?id=102704939189&ref=appd
9
A successful campaign has been launched by the utility SEAS-NVE, see
www.maalerjagten.dk
10
DIY Kyoto, www.diykyoto.com/uk/wattson/about
11
Onzo Ltd., www.onzo.co.uk
12
Current Cost, www.currentcost.com
13
Blue Line Innovations Inc., www.bluelineinnovations.com
14
Energy Inc., www.theenergydetective.com
4
5. mainly for safety reasons, the wiring around household meters is in-
accessible in many countries and modifications require a technician.
Another drawback is that they are unsuitable for providing users
with feedback on the consumption of individual devices which,
from a feedback perspective, would be necessary to draw conclu-
sions on how consumption and behavior relate to each other.
Some experimental systems attempt to disaggregate the total con-
sumption measured by a single sensor to provide more specific in-
formation about electricity consumption at device level [28]. The
aim of these non-intrusive load monitoring systems is to keep
equipment costs and installation effort to a minimum, but still ob-
tain detailed energy usage data. To determine which appliances
are currently running, some of these systems simply measure the
overall power difference from one point in time to the next; a
principle that has been investigated by several researchers in the
past [2], [9], [37]. Other more sophisticated approaches use statis-
tical signature analysis and pattern detection algorithms to infer
the devices in question from the current and voltage wave forms
[23]. To achieve disaggregation, these systems require either a
priori knowledge about the household devices and their electrical
characteristics, or entail a complex calibration and training phase
involving the user, in which the system learns about specific de-
vice characteristics. However, a priori knowledge is difficult to
obtain in a world of fast-changing small appliances, and manual
training is a significant barrier to usage. Furthermore, appliances
whose power consumption varies or overlaps with that of other
devices pose a particular challenge for disaggregation algorithms.
A more sophisticated idea has been explored by Patel et al. [35].
The authors developed a system that relies on a single sensor that
can be plugged in anywhere on a domestic electric circuit. On the
residential power line, it listens for unique noise changes caused
by the abrupt switching of devices. This approach makes it possi-
ble to determine, with some probability, the status (such as on, off,
stand-by, etc.) of an appliance. To infer the actual electricity con-
sumption of a device, this information then has to be combined
with the measurements of a smart meter.
4.2 Multiple Sensor Approach
Multiple sensor approaches can be subdivided into direct and indi-
rect systems. Direct systems require an in-line sensor to be in-
stalled for every device or circuit. Indirect sensing systems use a
central electricity meter together with additional context sensors
to monitor energy consumption.
Direct sensing systems mostly come in the form of smart power
outlets. They are relatively easy to deploy and several products
exist15
. Once installed, they measure the attached load and display
the measurement data on the unit itself or transmit it wirelessly to
a remote display. However, these systems are not able to aggre-
gate consumption from multiple sensors and combine the different
data to form a comprehensive picture.
To overcome this limitation, other work has focused on developing
systems that combine multiple power sensors. Guinard et al. [13]
realized a system that integrates smart power sockets (“Ploggs”16
)
which communicate their measurements via Bluetooth or Zigbee.
A gateway is responsible for identifying smart sockets that are
within range. It also makes their functionality available as re-
sources on the Web and provides local aggregation of device-level
15
For example “Kill a Watt”, www.p3international.com/products/special/
P4400/P4400-CE.htm
16
Plogg, www.plogginternational.com
services (e.g., the accumulated consumption of all sockets). Jiang
et al. [20] developed a system where sensors measure power con-
sumption at outlets and communicate their readings over a wire-
less IPv6 network to a server that populates a central database.
Multiple direct sensing systems all suffer from the fact that dep-
loying a large number of electricity sensors (i.e., meters) through-
out a house quickly becomes expensive. Indirect sensing systems
try to remedy this by keeping intrusion into the electrical system
at a minimum. Instead of many power meters, they use other types
of context sensors. In [21] Kim et al. describe a system that uses a
single electrical sensor to measure the entire electricity consump-
tion of a household together with additional context sensors (such
as light, sound, and electromagnetic sensors) that help to infer
which appliance is currently operating from the measurable sig-
nals it emits. Within a defined set of appliances, the authors show
that the system can estimate device-level power consumption within
a 10% error range. However, the system’s performance depends
greatly on the correct calibration and placement of the distributed
context sensors, which is not an easy task for the average user.
4.3 Feedback: Characterization and Outlook
Table 2 summarizes the main advantages and disadvantages of the
various electricity feedback systems. Single sensor systems are
hard to deploy but reasonably priced, and once installed they have
a low usage barrier. Since the single sensor is typically installed
close to the household meter or in the fuse box, overall consump-
tion is easy to monitor. However, to obtain information at device
level calls for more sophisticated approaches that require algo-
rithms to be calibrated. In addition, due to the wide variety of
electrical devices involved, the accuracy of these systems is
somewhat limited.
Table 2. Properties of different energy monitoring solutions
Character-
istics
Single
sensor
Multiple sensors
Direct in-line Indirect
Installation Hard Medium Hard
Cost Low High High
Usage barrier Low High High
Calibration Hard Easy Hard
Device level
accuracy
Low High Medium
Household
level accuracy
High Low High
In contrast, direct in-line electricity monitoring systems are very
accurate at device level since the electricity is measured at the de-
vice itself. However, this advantage comes at a high cost, as in
principle every appliance has to be equipped with a sensor. At the
same time this increases the usage barrier, since most users are not
willing to install a large number of sensors or smart power outlets
throughout the house. Therefore such systems will typically only
cover a subset of all electricity consuming devices in a household.
Finally, indirect systems are theoretically able to provide feedback
both on overall electricity consumption and, to a certain extent, on
device-level electricity usage. However, they require users to deploy
various context sensors in the right places and necessitate complex
calibration, which leads to both high costs and a high usage barrier.
The way forward for electricity monitoring systems involves a
scenario in which household appliances, which today have only
5
6. limited capabilities, become more powerful and smart. Through
the integration of small, inexpensive embedded ICT components,
they would sense and transmit their current energy usage together
with other status information. Within the house, appliances could
communicate with each other (and with the smart meter) via an
established protocol (e.g., powerline, Zigbee, WLAN), although
dedicated new technologies such as digitalSTROM17
, rivaling tra-
ditional domestic network technologies (BACnet, EIB, KNX,
etc.), might also feature.
Moreover, the cost of integrating embedded Web servers (based
on REST and IPv6 / 6LowPAN) into household appliances should
in future be low. This would lead to a wide variety of application
scenarios in which the smart domestic electricity meter (or a simi-
lar device) could serve as a central component for data aggrega-
tion and analysis. At the same time, embedding a Web interface
into appliances would enable them to be fully integrated into the
Internet [13]. As well as the allocation of a device-specific Web
page for status information, this would allow the device to be con-
trolled and its data to be processed using the full power of Web
2.0 tools, giving rise to a “Web of Things” [12]. It is obvious,
however, that with such possibilities we would need to pay se-
rious attention to privacy and security issues.
5. THE eMETER SYSTEM
In this section, we present the eMeter system that is based on a
single sensor approach and attempts to overcome most of the limi-
tations described above. By connecting a smart electricity meter
with a mobile phone application, the system is particularly easy to
use and realizes those features that seem to be most promising in
terms of energy feedback. According to the literature [11], effec-
tive energy feedback has to
− feature a low usage barrier,
− be presented on a device that is already integrated into users’
daily life,
− be given frequently, in real time, and be available when
needed, and
− provide the ability to apportion total electricity consumption.
The eMeter system considers these issues. It achieves a low usage
barrier by using a smart electricity meter, which is going to be in-
stalled in households throughout Europe by law anyhow. So all that
users have to do is install a mobile phone application that can easily
be downloaded from the Internet. The system is simple to setup
and requires no modification by the user – neither to the electrical
wiring, nor by deploying additional hardware at device level [46].
By providing real-time feedback on a mobile phone, the system
features not only feedback on a device that is already part of the
user’s life, but also the ability to provide instantaneous feedback
that is available when needed. This is especially important since
trials have shown that when using an additional battery-powered
display for electricity feedback, 50% of all users do not replace
the battery once it is depleted [40]. This indicates a loss of interest
after the users’ initial curiosity has been satisfied. Since they are
not integrated into users’ daily life, these additional displays do
not seem capable of motivating users for long periods of time.
Lastly, useful feedback has to link specific actions to their effects by
providing the ability to disaggregate overall electricity consumption.
In order to take effective measures, it is vital to understand how
much power individual devices consume in standby mode or while
17
www.digitalstrom.org
operating [36]. The eMeter’s interactive measurement functionality
allows users to measure the consumption of almost every device
that can be manually switched on or off (see Section 5.2 below).
5.1 The eMeter Architecture
The eMeter system consists of three independent components
(Figure 1): a smart electricity meter that monitors the total domes-
tic load; a gateway that manages and provides access to the
logged measurement data; and a portable user interface on a mo-
bile phone that provides real-time feedback on energy consump-
tion and enables users to interactively monitor, measure, and
compare their energy consumption.
The system architecture is based on the REST (Representational
State Transfer) paradigm [10]. REST is a resource-oriented ap-
proach that enables physical resources to be easily and seamlessly
integrated into the Web. For this purpose, REST proposes two ba-
sic principles. First, transferring the conventional operation-
centric model view into a data-centric view, which essentially
means that services now become resources that can be identified
and manipulated (i.e., transferred, indexed, put on Web pages etc.)
by using URLs. Second, the only available operations to access,
update, delete, and create resources are the four main operations
provided by HTTP (GET, POST, DELETE, PUT).
The first component of the architecture is the smart electricity me-
ter (provided by Landis+Gyr in our implementation). It logs the
load induced by all the devices attached to the residential power
line. In contrast to traditional electricity meters, the smart meter
has a communication interface for remote meter readings (typical-
ly used by the energy utility company). In order to achieve real-
time feedback, we exploit this functionality by asking the meter to
transmit all available data via its interface every second.
The second component, a lightweight gateway, is implemented in
Java and consists of a parser, a database, and a small Web server
(based on the RECESS!18
framework). In order to continuously
acquire the logged data from the smart meter in near real time, the
integrated SML19
parser automatically polls the meter every
second and stores the data it receives in an SQL database. Access
to the gateway’s functionality and also to the smart meter data is
provided by the Web server using URLs.
The smart meter measures a number of different physical values
(e.g., actual load, voltage, current, etc.). Through the gateway, they
all become hierarchically structured resources in the sense of REST.
That is, each of the resources implements the four basic HTTP
verbs. This is a powerful concept since it allows the meter data to
be accessed via any Web browser. For example, just by calling
http://serverAddress]/emeter/energyServer/
smartMeter/1/measurements.json?c=last
18
www.recessframework.org
19
Smart Message Language, www.t-l-z.org/docs/SML_080711_102_eng.pdf
Figure 1. Smart meter communicating with the mobile UI
6
7. the resource measurement can be monitored. The corresponding
GET request issued by the Web browser is answered by the gate-
way, which first routes the request to the resource (that takes care
of reading the “last” value) and then wraps the result in the form
of an HTTP or JSON message as shown below.
{
"smartMeter":
{"id":"1","name":"Landis+Gyr","createdOn":1248102873},
"measurements":
{"id":"9513463","date":1261401851,"watts":322.483}
}
Since the gateway can support multiple formats, we decided to
use JSON (as a lightweight alternative to XML) for interaction
with other applications, and HTML for providing a human-read-
able representation in a Web browser.
The third component, the content-rich user interface on a mobile
phone, is implemented using Objective-C. It exploits the function-
ality provided by the gateway to access the meter readings and dy-
namically visualize the information in real time. It does this by
calling appropriate URLs on the gateway together with the cor-
responding HTTP verb, and processing the JSON message it rece-
ives in response. The user interface is also responsible for trans-
mitting user-generated data, such as details about the household
and appliances, to the gateway.
The architecture we have described here is not restricted to the
eMeter system. It shows in general how systems for home auto-
mation and similar tasks can be designed to provide detailed, real-
time feedback. It also shows that directly integrating smart physi-
cal objects into the Web infrastructure greatly facilitates the de-
velopment of applications (such as the mobile user interface in our
case). This is prototypical for an emerging concept known as the
“Web of Things” [12].
5.2 The eMeter User Interface
In order to provide the important feedback features mentioned
earlier, the eMeter user interface consists of the following four
views (Figure 3): Live visualization of current electricity con-
sumption (a), a historical view of energy consumption (b, c), a de-
vice inventory view that displays energy usage and costs per
measured device (d), and a measurement view (e) which enables
the user to interactively measure the consumption of almost any
switchable electrical appliance in the house.
The current consumption view (Figure 3a) shows current consump-
tion in real-time. The color-coded self-learning scale allows users to
assess how their current consumption compares to their historical
consumption readings (green to red). The blue part of the scale de-
picts the level of standby electricity consumption in the home.
The history view (Figure 3b, 3c) shows a line chart of historical
consumption. Users can choose between different time periods,
e.g. previous hour, previous day, etc. Together with the chart, this
view displays equivalents such as kWh and cost for the accumu-
lated consumption over the last five selected periods (Figure 3b,
lower part). The color-coded bars allow users to compare their
historical consumption to that of a typical average household of
the same size in the same location. The historical consumption
view also provides budget calculations and projections (Figure 3c).
The device inventory view (Figure 3d) lists all previously meas-
ured devices. In addition, it allows users to view device details
and assign a location (e.g., a room) as well as a particular utiliza-
tion scheme (upon which the device’s cost calculations are based)
to the device. It also enables users to sort the readings by location
or the amount of power used, so that the biggest energy guzzler
appears at the top.
The measurement view (Figure 3e) enables users to interactively
measure the electricity consumption of most switchable appli-
ances in the household. To perform a measurement, the user simp-
ly activates the process by pressing the green start button and then
turns the device being measured on or off. The corresponding re-
sult is shown on the display within seconds (Figure 2). The neces-
sary calculations for this are performed on the mobile phone – as
soon as the user initiates the measurement, the current consump-
tion value determined by the smart meter is stored, and the meas-
urement algorithm on the phone then waits for a significant
change in this value. It then calculates the difference between the
Figure 2. User measuring the power consumption of devices
a b c d e
Figure 3. eMeter user interface (from left to right):
current consumption view, history view (aggreg. consumption), history view (budgeting), device inventory view, measurement view
7
8. two values. (Incidentally, if another device is switched on or off
during the measurement interval, the result may be incorrect.
However, because this generates a spurious reading, users are typ-
ically aware of the situation and can simply repeat the process.)
After the measurement has been carried out, users can save the
measured device to a list of appliances. The user interface pro-
vides additional options for personalization. For example, users
can take photos of the measured appliance, or detail its utilization
to calculate the annual costs incurred. If a device category is se-
lected, the user interface displays category-specific energy effi-
ciency information and guidance on how to save energy.
5.3 eMeter: Summary and Future Work
The eMeter system allows users to interactively monitor, measure,
and compare their energy consumption at a household and device
level. The system attempts to overcome the discouraging installa-
tion overheads incurred by other typical energy feedback systems
by making use of a smart electricity meter, whose installation is be-
coming mandatory in Europe and which provides highly accurate
readings at a household level. Assuming that the lightweight gate-
way component will in future be integrated into the smart meter it-
self (or into another suitable device such as a DSL router or an Inter-
net gateway), users would only have to download and install the mo-
bile phone application – something that can be done easily in minutes
or even seconds. Since the system requires no additional hardware, it
is inexpensive and should generally have a low usage barrier.
The system’s accuracy at device level, however, suffers from its
single sensor approach. We are trying to overcome this by inte-
grating into the user interface measurement functionality aimed at
providing users with an initial idea of how much energy different
devices consume. Provided the brief measurement interval is rep-
resentative, it enables users to measure the consumption of any
switchable or pluggable device. The only real challenge might be
that of correctly measuring devices whose energy consumption is
variable or dynamic (such as laptops). However, some household
devices that consume a non-negligible amount of energy, such as
washing machines and freezers, cannot usually just be turned on
or off. For such devices that cannot easily be measured by the us-
er, one of the automatic device identification methods described in
Section 4 could be envisaged.
6. DISCUSSION
ICT can make a significant contribution to saving energy, both
through autonomous optimization and by inducing changes in user
behavior. But achieving the latter is not that easy. Bearing in mind
the somewhat disappointing results from early feedback systems
using smart metering infrastructures, one could indeed question
how effective getting the “user into the loop” really is. However,
the rather unsatisfactory results seem in the most part to be a con-
sequence of insufficient ways of motivating and engaging con-
sumers, as notable efficiency gains have been demonstrated in
many of the better organized settings (see Table 1).
Even if the direct savings generated by feedback systems are only
in the order of a few percent, the cheapest and most environmen-
tally-friendly type of energy is still that which is not produced in
the first place because it is not needed. Moreover, society should
also benefit indirectly from the “user in the loop” paradigm. A
raised awareness of energy consumption is not only expected to
lead to improved usage patterns, but also to an increased willing-
ness to pay a premium for energy-efficient goods and services.
These spillover effects (people who frequently deal with con-
sumption information are more likely to consider environmental
aspects when purchasing a new TV, and are likely to choose an
energy-efficient car) can help to achieve additional savings.
The use of “smart” ICT for sustainability brings many other issues
to the fore. For example, integrating smart cooperating real-world
objects into environments other than energy management systems
is an interesting field of research, which goes hand in hand with
the growing number of Web interfaces and Internet-enabled de-
vices around us. This development might accelerate the emer-
gence of a so-called Internet of Things [27].
Security is another crucial issue. Smart meters for example, which
often not only measure and communicate consumption readings
but are also able to remotely reduce the load or disconnect house-
holds from the power grid, become critical infrastructure compo-
nents. A virus that caused devices to malfunction or a denial of
service attack could lead to serious damage. Moreover, the elec-
tricity infrastructure is intended to be a long-term undertaking, but
network security concepts and methods such as key lengths and
encryption algorithms – and the possibilities open to attackers –
change at a much faster pace.
Usability and reliability are also important. Even people who are
totally unfamiliar with computers or network security now have a
networked computer in their home in the form of a smart meter –
rebooting it by hand, manually updating the device, or dealing
with cryptic error codes is not an option.
Also, privacy concerns are often raised, especially in the context
of smart metering. In fact, detailed knowledge of a household’s
electrical device usage may reveal much about the habits of its
occupants.20
Leaving most consumption data inside the house and
only transferring data that is essential for billing might be part of
the solution. However, this rules out some interesting global op-
timizations and remote services that require detailed real-time
energy consumption data. Also, convincing people to trust in the
protective approach might turn out to be a challenge.
Since advanced metering makes fine-grained energy consumption
data available, this raises the question of how to exploit this data
to develop valuable services that improve energy efficiency. This
issue has also recently attracted the attention of industry giants
such as Google21
and Microsoft22
, who might be on their way to
becoming service providers (e.g., for automatic energy consulting)
in the residential energy sector. Data analytics and pattern recog-
nition algorithms are essential for such services, and might then
help consumers to conserve energy (or at least understand their
electricity bills better…).
The coalescence of the Internet of Things and energy topics will
also promote the development of new product-as-a-service con-
cepts, and provide new stimuli for adopting home automation sys-
tems. It will thus also strengthen interest in business service re-
search in a sector that so far has limited experience in dealing with
private users.
When it comes to influencing consumer behavior, further research
is required not only to develop user interfaces that present con-
20
In their analysis [24], Lisovich and Wicker come to the conclusion that
increased availability of data, along with emerging use cases, will inevi-
tably create or exacerbate issues of privacy and that there are strong mo-
tivations for entities involved in law enforcement, advertising, and crim-
inal enterprises to collect and repurpose power consumption data.
21
Google Power Meter: www.google.org/powermeter
22
Microsoft Hohm: www.microsoft-hohm.com
8
9. sumption data in a suitable way, but also to identify and better un-
derstand concepts from behavioral science such as framing, goal
setting, and identity signaling and their potential to induce sus-
tainable change. Moreover, it is important to identify engagement
strategies (e.g., games, competitions, and rewards) that help to
boost consumer involvement once their initial curiosity has been
satisfied. For these purposes, ICT is not only a means of imple-
mentation, but also a research tool that allows the effects of such
measures to be observed in a timely and precise way.
Further research is also necessary to quantify or qualify the effi-
ciency gains and energy savings that can be attained by ICT
usage. In an absolute setting, for example, it is difficult to deter-
mine how much energy smart metering can conserve. Results re-
ported from pilot studies are only valid for the specific application
domain, the technology under consideration, the user group, and
other contextual conditions such as accompanying campaigns.
Spillover and other indirect effects make an assessment even more
difficult.
ICT’s indirect consequences on energy consumption are particu-
larly difficult to analyze. On the negative side, one has to consider
so-called rebound effects – someone with an energy-efficient car,
for example, might partly compensate for the savings achieved by
technology by simply driving more, because it is now cheaper.
Some researchers are even warning that there is some risk that
ICT will become counterproductive with regard to general envi-
ronmental sustainability, or that it has only a low overall effect
because positive and negative environmental impacts partially
cancel each other out when aggregated [17]. For example, one
important factor relating to increased energy consumption is
demonstrated by ICT applications that make freight and passenger
transport more efficient (i.e., cheaper or faster), because this
creates more traffic and thus possibly induces more energy (i.e.,
fuel) consumption. In a thorough study on the rebound effect [43],
Sorell concludes that this effect has generally been neglected
when assessing the potential impact of energy efficiency policies.
Analyzing and mitigating such opposing effects should therefore
be a focal point of future research.
On the other hand, ICT exerts a major influence by enabling energy
efficiencies in other sectors (such as logistics, transportation, and
building infrastructure). Buildings, for example, account for 40% of
the EU’s energy requirements, and it is estimated that almost 35%
of the energy used in the residential buildings sector could be saved
by 2020 [8]. Some even expect that ICT’s potential to help other
sectors become more energy efficient could deliver greenhouse
gas emission savings five times greater than ICT’s own foot-
print23
. Furthermore, ICT enables a shift from material goods to
services and promotes a general structural change towards a less
material-intensive economy. While the long-term consequences of
dematerialization are difficult to predict, one can at least hope that
in total it should have a beneficial effect on sustainability.
With all that, however, one must not forget that ICT has its own en-
vironmental footprint. ICT components not only consume energy,
but their fabrication and disposal is also an important factor to be
taken into consideration. The environmental effects of the laborious
mining, processing, and usage of rare materials (such as tantalum,
indium, niobium, etc.) to build these components must also be con-
sidered. Advances in technology and its application should not de-
tract us from the numerous problems relating to obtaining and recy-
cling the basic materials that are used to construct ICT systems [16].
23
www.smart2020.org
7. OUTLOOK – GREEN ICT FOR GREEN
Clearly, if we are to achieve an economy based on sustainable
energy, “ICT for green” is still in need of much work. We need to
develop not only feedback systems as described above, but also
large-scale distributed energy management systems that can deal
with huge amounts of event data and operate in real time, as well
as infrastructures such as electronic market platforms that support
the cooperation of various players and thus help to automatically
balance highly fluctuating energy supply and demand. And of
course these systems have to be reliable, secure, and cost-
effective.
Despite these and all the other challenges mentioned above, we
are convinced that ICT, when used in a “smart” way, will signifi-
cantly help to reduce society’s demand for carbon-based energy,
while at the same time offering interesting business opportunities
for industry and guaranteeing a desirable lifestyle for the citizens.
It should be clear that “green ICT” and “ICT for green” are not
mutually exclusive – both are important, and they complement
each other [5]. Hence the challenge for the future lies in the ap-
pealing synthesis “green ICT for green”.
8. REFERENCES
[1] A. Arvola, A. Uutela, and U. Anttila. Billing feedback as a
means of encouraging conservation of electricity in house-
holds: a field experiment in Helsinki. In: A. Arvola, E.
Rautavaara, and A. Uutela (eds): Energy and the consumer.
Final report on the research program 1990-1992, 46–60,
Ministry of Trade and Industry, Helsinki, Finland, 1994.
[2] M. Baranski, J. Voss. Genetic algorithm for pattern detec-
tion in NIALM systems. Proc. IEEE Int. Conf. on Systems,
Man and Cybernetics, Vol 4, 3462–3468, The Hague, the
Netherlands, 2004.
[3] M. Chetty, D. Tran, and R. Grinter. Getting to green: under-
standing resource consumption in the home. Proc.
Ubicomp08, 242–251, Seoul, South Korea, 2008.
[4] Commission of the European Communities. Commission
Recommendation on mobilising Information and Commu-
nications Technologies to facilitate the transition to an
energy-efficient, low-carbon economy. C(2009) 7604 final,
Brussels, Belgium, 2009.
[5] V. Coroama, L. Hilty. Energy consumed vs. energy saved
by ICT – A closer look. Proc. EnviroInfo 2009, 353–361,
Berlin, Germany, 2009.
[6] S. Darby. The effectiveness of feedback on energy con-
sumption. A Review for DEFRA of the Literature on
Metering, Billing and Direct Displays, 2006.
[7] K. Davis. Prepaid Metering Can Bring Strays Back to Your
Fold. Utility Automation and Engineering T&D, May 2003.
[8] European Commission DG INFSO. Impacts of Information and
Communication Technologies on Energy Efficiency. 2008.
[9] L. Farinaccio, R. Zmeureanu. Using a pattern recognition
approach to disaggregate the total electricity consumption in
a house into the major end-uses. Energy and Buildings,
1999, 30(3):245–259.
[10] R. Fielding, R. Taylor. Principled Design of the Modern
Web Architecture. ACM Trans. Internet Technology, 2002,
2(2):115–150.
9
10. [11] C. Fischer. Feedback on household electricity consumption: a
tool for saving energy? Energy Efficiency, 2008, 1(1):79–104.
[12] D. Guinard, V. Trifa, and E. Wilde. Architecting a Mashable
Open World Wide Web of Things. TR ETH Zürich, 2010.
[13] D. Guinard, M. Weiss, and V. Trifa. Are you Energy-
Efficient? Sense it on the Web! Adjunct Proc. Pervasive09,
Nara, Japan, 2009.
[14] O. Günther. Environmental information systems. SIGMOD
Rec., 1997, 3(1):3–4.
[15] J. Henryson, T. Hakansson, and J. Pyrko. Energy efficiency
in buildings through information – Swedish perspective.
Energy Policy, 2000, 28(3):169–180.
[16] L. Hilty. Electronic waste – an emerging risk? Environmen-
tal Impact Assessment Review. 25(5):431–435, 2005.
[17] L. Hilty, P. Arnfalk, L. Erdmann, J. Goodman, M. Leh-
mann, and P.A. Wäger. The relevance of information and
communication technologies for environmental sustainabili-
ty – a prospective simulation study. Environmental Model-
ling & Software, 2006, 21(11):1618–1629.
[18] A. Ilic, P. Bowman, J. Ng, and T. Staake. The Value of
RFID for RTI Management. Electronic Markets, 2009,
19(2–3):125–135.
[19] International Energy Agency. World Energy Outlook 2009.
OECD Publishing, 2009.
[20] X. Jiang, S. Dawson-Haggerty, P. Dutta, and D. Culler. De-
sign and Implementation of a High-Fidelity AC Metering Net-
work. Proc. IPSN09, 253–264, San Francisco, USA, 2009.
[21] Y. Kim, T. Schmid, Z. Charbiwala, and M. Srivastava.
ViridiScope: Design and Implementation of a Fine Grained
Power Monitoring System for Homes. Proc. Ubicomp09,
245–254, Orlando, USA, 2009.
[22] J.A. Laitner, K. Ehrhardt-Martinez. Information and Com-
munication Technologies: The Power of Productivity. Envi-
ronmental Quality Management, 2009, Part I: 18(2):47–66,
Part II: 18(3):19–35.
[23] H. Lam, G. Fung, and W. Lee. A Novel Method to Construct
Taxonomy Electrical Appliances Based on Load Signatures.
IEEE Trans. Consumer Electronics, 2007, 53(2):653–660.
[24] M. Lisovich, S. Wicker. Privacy Concerns in Upcoming
Residential and Commercial Demand-Response Systems.
Proc. Power Systems Conference, Clemson, USA, 2008.
[25] E.A. Locke, G.P. Latham. Self–Regulation through Goal
Setting. Organizational Behavior and Human Decision Proc-
esses, 1991, 50(2):212–247.
[26] C. Loock, T. Staake, and E. Fleisch. Kundenportale in der
Energiebranche: Bestandsaufnahme und Entwicklungspo-
tenziale. Zeitschrift für Energiewirtschaft, 2009, 3:268–269
[27] F. Mattern, C. Flörkemeier. From the Internet of Computers
to the Internet of Things. Informatik-Spektrum, 2010, 33(2).
[28] H.S. Matthews, L. Soibelman, M. Berges, and E. Goldman.
Automatically Disaggregating the Total Electrical Load in
Residential Buildings: a Profile of the Required Solution.
Proc. Intelligent Computing in Engineering, 381–389, Ply-
mouth, UK, 2008.
[29] K.E. McShane, J.A. Cunningham. The role of relevancy in
normative feedback for university students’ drinking pat-
terns. Addictive Behaviors, 2003, 28(8):1523–1528.
[30] D. Mountain. The impact of real-time feedback on residential
electricity consumption: The Hydro One pilot. Mountain Eco-
nomic Consulting and Associates Inc., Ontario, Canada, 2006.
[31] L. Nielsen. How to get the birds in the bush into your hand:
Results from a Danish research project on electricity
savings. Energy Policy, 1993, 21(11):1133–1144.
[32] J. Nipkow, E. Bush. Stand-by-Verbrauch von Haushaltgerä-
ten. Bulletin SEV/VSE, 3/2004, 9–14.
[33] J. Nipkow, S. Gasser, and E. Bush. Der typische Haushalt-
Stromverbrauch. Bulletin SEV/VSE, 19/2007, 24–26.
[34] OECD Factbook 2008 – Economic, Environmental and
Social Statistics. OECD Publishing, 2008.
[35] S. Patel, T. Robertson, J. Kientz, M. Reynolds, and G.
Abowd. At the Flick of a Switch: Detecting and Classifying
Unique Electrical Events on the Residential Power Line.
Proc. Ubicomp07, 271–288, Innsbruck, Austria, 2007.
[36] J. Petersen, V. Shunturov, K. Janda, G. Platt, and K. Wein-
berger. Dormitory residents reduce electricity consumption
when exposed to real-time visual feedback and incentives. Int.
J. of Sustainability in Higher Education, 2007, 8(1):16–33.
[37] A. Prudenzi. A neuron nets based procedure for identifying
domestic appliances pattern-of-use from energy recordings
at meter panel. Proc. IEEE Power Engineering Society Win-
ter Meeting, New York, USA, 2002, 2:941–946.
[38] D. Pudjianto, C. Ramsay, and G. Strbac. Virtual Power
Plant and System Integration of Distributed Energy Re-
sources. Renewable Power Generation, 2007, 1(1):10–16.
[39] W.F. van Raaij, T.M.M. Verhallen. A behavioral model of
residential energy use. Journal of Economic Psychology,
1983, 3:39–63.
[40] L. van Rensburg. Energy demand research project: Review
of progress for period 04/2008 – 08/2008. TR Ofgem, 2009.
[41] P.W. Schultz, J.M. Nolan, R.B. Cialdini, N.B. Goldstein,
and V. Griskevikins. The constructive, destructive and re-
constructive power of social norms. Psychological Science,
2007, 18(5):429–434.
[42] D. Siegel. The Impact of Computers on Manufacturing Pro-
ductivity Growth: A Multiple-Indicators, Multiple-Causes
Approach. The Review of Economics and Statistics, 1997,
79(1):68–78.
[43] S. Sorell. The Rebound Effect: an Assessment of the Evi-
dence for Economy-wide Energy Savings from Improved
Energy Efficiency. UK Energy Research Centre, 2007.
[44] E.M. Tapia, S.S. Intille, and K. Larson. Activity Recogni-
tion in the Home Using Simple and Ubiquitous Sensors.
Proc. Pervasive 2004, 158–175, Vienna, Austria, 2004.
[45] U.S. Environmental Protection Administration. Inventory of
U.S. Greenhouse Gas Emissions and Sinks: 1990–2007. 2009.
[46] M. Weiss, T. Graml, T. Staake, F. Mattern, and E. Fleisch.
Handy feedback: Connecting smart meters with mobile
phones. Proc. MUM 2009, Cambridge, UK, 2009.
[47] H. Wilhite, R. Ling. Measured energy savings from a more
informative energy bill. Energy and Buildings, 1995, 22(2):
145–155.
10